Import and process data

Learning

Model: Correct responses by age, trial, block number, and block condition

	correct_response_made
Predictors	Odds Ratios	SE
age scaled	1.2779	0.0482
learning trial scaled	1.6933	0.0341
reward condition1	1.6760	0.0456
block number scaled	1.2739	0.0359
age scaled × learning trial scaled	1.1295	0.0228
age scaled × reward condition1	1.0923	0.0298
learning trial scaled × reward condition1	1.1689	0.0181
age scaled × block number scaled	0.9848	0.0280
learning trial scaled × block number scaled	1.0768	0.0177
reward condition1 × block number scaled	1.0705	0.0246
age scaled × learning trial scaled × reward condition1	1.0624	0.0167
(age scaled × learning trial scaled) × block number scaled	1.0138	0.0170
(age scaled × reward condition1) × block number scaled	0.9966	0.0232
(learning trial scaled × reward condition1) × block number scaled	1.0220	0.0159
(age scaled × learning trial scaled × reward condition1) × block number scaled	1.0148	0.0161
Random Effects
σ²	3.29
τ₀₀ _{subject_id}	0.17
τ₁₁ _{subject_id.re1.learning_trial_scaled}	0.02
τ₁₁ _{subject_id.re1.reward_condition1}	0.07
τ₁₁ _{subject_id.re1.block_number_scaled}	0.08
τ₁₁ _{subject_id.re1.learning_trial_scaled_by_reward_condition1}	0.00
τ₁₁ _{subject_id.re1.learning_trial_scaled_by_block_number_scaled}	0.01
τ₁₁ _{subject_id.re1.reward_condition1_by_block_number_scaled}	0.04
τ₁₁ _{subject_id.re1.learning_trial_scaled_by_reward_condition1_by_block_number_scaled}	0.00
ρ₀₁
ρ₀₁
ICC	0.05
N _{subject_id}	151
Observations	38352
Marginal R² / Conditional R²	0.161 / 0.203

Figure 2A: Correct response by block condition, stimulus repetition, and age group

Supplementary Figure: Correct response by block condition and block number

Figure 2B: Generalization by block condition, category repetition, age group

Model: Correct response to first appearance of each stimulus

	correct_response_made
Predictors	Odds Ratios	SE
age scaled	1.0717	0.0245
category rep scaled	1.2596	0.0295
reward condition1	1.7317	0.0458
block number scaled	1.0951	0.0246
age scaled × category rep scaled	1.0698	0.0252
age scaled × reward condition1	1.0792	0.0286
category rep scaled × reward condition1	1.3324	0.0312
age scaled × block number scaled	1.0003	0.0227
category rep scaled × block number scaled	1.0504	0.0245
reward condition1 × block number scaled	1.0957	0.0246
age scaled × category rep scaled × reward condition1	1.0881	0.0257
(age scaled × category rep scaled) × block number scaled	1.0288	0.0243
(age scaled × reward condition1) × block number scaled	1.0027	0.0227
(category rep scaled × reward condition1) × block number scaled	1.0722	0.0250
(age scaled × category rep scaled × reward condition1) × block number scaled	1.0346	0.0244
Random Effects
σ²	3.29
τ₀₀ _{subject_id}	0.01
τ₁₁ _{subject_id.re1.block_number_scaled}	0.01
τ₁₁ _{subject_id.re1.reward_condition1}	0.03
τ₁₁ _{subject_id.re1.block_number_scaled_by_reward_condition1}	0.01
ρ₀₁
ρ₀₁
ICC	0.00
N _{subject_id}	151
Observations	11267
Marginal R² / Conditional R²	0.116 / 0.118

Supplementary Figure: Generalization by trial, reward condition, and block number

Figure 2C: WSLS by age group

Supplementary Figure: WSLS by block number

Model: Category win-stay lose-shift

	WSLS
Predictors	Odds Ratios	SE
age scaled	1.1791	0.0378
learning trial scaled	1.1812	0.0180
reward condition1	2.4283	0.0719
block number scaled	1.1835	0.0243
age scaled × learning trial scaled	1.0659	0.0166
age scaled × reward condition1	1.1856	0.0351
learning trial scaled × reward condition1	1.3255	0.0202
age scaled × block number scaled	0.9832	0.0204
learning trial scaled × block number scaled	1.0061	0.0153
reward condition1 × block number scaled	1.1665	0.0181
age scaled × learning trial scaled × reward condition1	1.0783	0.0168
(age scaled × learning trial scaled) × block number scaled	1.0220	0.0158
(age scaled × reward condition1) × block number scaled	1.0060	0.0159
(learning trial scaled × reward condition1) × block number scaled	1.0192	0.0155
(age scaled × learning trial scaled × reward condition1) × block number scaled	0.9944	0.0154
Random Effects
σ²	3.29
τ₀₀ _{subject_id}	0.11
τ₁₁ _{subject_id.re1.reward_condition1}	0.09
τ₁₁ _{subject_id.re1.block_number_scaled}	0.03
ρ₀₁
ρ₀₁
ICC	0.03
N _{subject_id}	151
Observations	35986
Marginal R² / Conditional R²	0.216 / 0.243

Stats: WSLS in first 10 trials

reward_condition	mean_WSLS	se_WSLS
Category-Predictive	0.5956	0.01045
Exemplar-Predictive	0.5899	0.01063

Memory

Memory delay stats

mean_delay	sd_delay	min_delay	max_delay
7.132	1.3	6	10

age_group	mean_delay	sd_delay	min_delay	max_delay
Children	7.2	1.325	6	10
Adolescents	7.38	1.323	6	10
Adults	6.824	1.212	6	10

Supplementary Figure: AUC values by stimulus repetitions, block condition, age group

Model: AUC values by stimulus repetitions, block condition, age

	AUC
Predictors	Estimates	SE
age scaled	0.01	0.01
reward condition1	-0.01	0.00
stim rep scaled	0.04	0.00
age scaled × reward condition1	-0.00	0.00
age scaled × stim rep scaled	-0.00	0.00
reward condition1 × stim rep scaled	0.00	0.00
(age scaled × reward condition1) × stim rep scaled	0.00	0.00
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.60
N _{subject_id}	151
Observations	912
Marginal R² / Conditional R²	0.162 / 0.662

Figure 3A: AUC values by age group, memory specificity, block condition

Model: AUCs by age, reward condition, memory specificity

	AUC
Predictors	Estimates	SE
age scaled	0.0159	0.0069
reward condition1	-0.0084	0.0024
foil type1	0.0600	0.0024
age scaled × reward condition1	-0.0042	0.0024
age scaled × foil type1	0.0021	0.0024
reward condition1 × foil type1	0.0013	0.0024
age scaled × reward condition1 × foil type1	0.0016	0.0024
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.63
N _{subject_id}	151
Observations	608
Marginal R² / Conditional R²	0.286 / 0.739

Model: AUCs by age, reward condition, foil type, memory delay

	AUC
Predictors	Estimates	SE
age scaled	0.0123	0.0069
reward condition1	-0.0083	0.0025
foil type1	0.0598	0.0025
memory delay scaled	-0.0148	0.0068
age scaled × reward condition1	-0.0038	0.0025
age scaled × foil type1	0.0016	0.0025
reward condition1 × foil type1	0.0014	0.0025
age scaled × memory delay scaled	-0.0137	0.0069
reward condition1 × memory delay scaled	0.0021	0.0025
foil type1 × memory delay scaled	-0.0026	0.0025
age scaled × reward condition1 × foil type1	0.0019	0.0025
(age scaled × reward condition1) × memory delay scaled	0.0013	0.0025
(age scaled × foil type1) × memory delay scaled	-0.0020	0.0025
(reward condition1 × foil type1) × memory delay scaled	-0.0001	0.0025
(age scaled × reward condition1 × foil type1) × memory delay scaled	0.0021	0.0025
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.62
N _{subject_id}	151
Observations	604
Marginal R² / Conditional R²	0.312 / 0.738

Supplementary Figure: AUC values with delay

RL modeling

Choice weights

Figure 2E: Distribution of choice weights

Model: Choice weights by block condition and age

	est
Predictors	Estimates	SE
abstraction1	-0.4633	0.0371
reward condition1	0.2448	0.0371
age scaled	0.1835	0.0507
abstraction1 × reward condition1	0.3241	0.0371
abstraction1 × age scaled	-0.0723	0.0371
reward condition1 × age scaled	0.0542	0.0371
(abstraction1 × reward condition1) × age scaled	0.0567	0.0371
Random Effects
σ²	0.83
τ₀₀ _{subject_id}	0.18
ICC	0.18
N _{subject_id}	151
Observations	604
Marginal R² / Conditional R²	0.296 / 0.422

Model: Exemplar choice weights across conditions

	est
Predictors	Estimates	SE
reward condition1	-0.0794	0.0421
Random Effects
σ²	0.53
τ₀₀ _{subject_id}	0.59
ICC	0.52
N _{subject_id}	151
Observations	302
Marginal R² / Conditional R²	0.006 / 0.526

Model: Category choice weights across conditions

	est
Predictors	Estimates	SE
reward condition1	0.5689	0.0509
Random Effects
σ²	0.78
τ₀₀ _{subject_id}	0.20
ICC	0.20
N _{subject_id}	151
Observations	302
Marginal R² / Conditional R²	0.250 / 0.400

Relations between choice weights and points earned

## Mixed Model Anova Table (Type 3 tests, S-method)
## 
## Model: total_points ~ age_scaled * beta_scaled * abstraction * reward_condition + 
## Model:     (1 | subject_id)
## Data: beta_ests_points
##                                                 Effect        df         F
## 1                                           age_scaled 1, 178.00 24.91 ***
## 2                                          beta_scaled 1, 529.48 19.81 ***
## 3                                          abstraction 1, 462.63      0.29
## 4                                     reward_condition 1, 447.92      0.56
## 5                               age_scaled:beta_scaled 1, 533.11    4.66 *
## 6                               age_scaled:abstraction 1, 463.17      1.14
## 7                              beta_scaled:abstraction 1, 499.31      1.72
## 8                          age_scaled:reward_condition 1, 448.06      0.17
## 9                         beta_scaled:reward_condition 1, 459.48      0.98
## 10                        abstraction:reward_condition 1, 455.16      1.90
## 11                  age_scaled:beta_scaled:abstraction 1, 506.72      0.31
## 12             age_scaled:beta_scaled:reward_condition 1, 456.20      0.84
## 13             age_scaled:abstraction:reward_condition 1, 452.91      0.15
## 14            beta_scaled:abstraction:reward_condition 1, 472.29 46.96 ***
## 15 age_scaled:beta_scaled:abstraction:reward_condition 1, 471.98      0.01
##    p.value
## 1    <.001
## 2    <.001
## 3     .591
## 4     .455
## 5     .031
## 6     .286
## 7     .190
## 8     .681
## 9     .322
## 10    .169
## 11    .576
## 12    .360
## 13    .700
## 14   <.001
## 15    .906
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '+' 0.1 ' ' 1

## 
## Call:
## lm(formula = total_points ~ est, data = beta_ests_points_c_c)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -279.190  -35.267    4.625   38.436  106.329 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  101.430      7.660  13.241  < 2e-16 ***
## est           38.762      4.837   8.013 2.99e-13 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 56.58 on 149 degrees of freedom
## Multiple R-squared:  0.3012, Adjusted R-squared:  0.2965 
## F-statistic: 64.21 on 1 and 149 DF,  p-value: 2.986e-13

## 
## Call:
## lm(formula = total_points ~ est, data = beta_ests_points_e_c)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -267.742  -34.803    7.474   43.234  123.277 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   133.74       8.64  15.480   <2e-16 ***
## est            10.84       4.37   2.482   0.0142 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 66.33 on 149 degrees of freedom
## Multiple R-squared:  0.03969,    Adjusted R-squared:  0.03324 
## F-statistic: 6.158 on 1 and 149 DF,  p-value: 0.01419

## 
## Call:
## lm(formula = total_points ~ est, data = beta_ests_points_c_e)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -324.49  -44.22    2.05   55.39  193.40 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  158.021      7.093   22.28   <2e-16 ***
## est           -6.648      6.924   -0.96    0.339    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 86.48 on 149 degrees of freedom
## Multiple R-squared:  0.006149,   Adjusted R-squared:  -0.0005209 
## F-statistic: 0.9219 on 1 and 149 DF,  p-value: 0.3385

## 
## Call:
## lm(formula = total_points ~ est, data = beta_ests_points_e_e)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -304.605  -36.510   -4.854   48.317  219.487 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   56.772     13.106   4.332 2.71e-05 ***
## est           58.972      6.908   8.537 1.47e-14 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 71.08 on 149 degrees of freedom
## Multiple R-squared:  0.3285, Adjusted R-squared:  0.324 
## F-statistic: 72.88 on 1 and 149 DF,  p-value: 1.468e-14

Figure 2D: Effect of choice weights on points earned

Relations between learning and memory

Do points earned during learning relate to memory?

Model: AUC by points earned

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: AUC ~ age_scaled * points_scaled * abstraction * reward_condition +  
##     (1 | subject_id)
##    Data: data
## 
## REML criterion at convergence: -1249
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.4573 -0.5178  0.0325  0.6283  2.5841 
## 
## Random effects:
##  Groups     Name        Variance Std.Dev.
##  subject_id (Intercept) 0.005858 0.07654 
##  Residual               0.003619 0.06016 
## Number of obs: 608, groups:  subject_id, 151
## 
## Fixed effects:
##                                                           Estimate Std. Error
## (Intercept)                                              7.710e-01  6.843e-03
## age_scaled                                               1.277e-02  6.865e-03
## points_scaled                                            8.960e-03  4.357e-03
## abstraction1                                             6.032e-02  2.607e-03
## reward_condition1                                       -6.419e-03  2.675e-03
## age_scaled:points_scaled                                -3.026e-04  4.321e-03
## age_scaled:abstraction1                                  4.972e-04  2.655e-03
## points_scaled:abstraction1                               5.044e-03  2.941e-03
## age_scaled:reward_condition1                            -2.624e-03  2.755e-03
## points_scaled:reward_condition1                         -3.825e-03  3.593e-03
## abstraction1:reward_condition1                          -1.082e-04  2.607e-03
## age_scaled:points_scaled:abstraction1                    1.693e-04  2.780e-03
## age_scaled:points_scaled:reward_condition1              -4.781e-03  3.312e-03
## age_scaled:abstraction1:reward_condition1               -9.863e-04  2.655e-03
## points_scaled:abstraction1:reward_condition1             6.957e-03  2.941e-03
## age_scaled:points_scaled:abstraction1:reward_condition1  4.803e-03  2.780e-03
##                                                                 df t value
## (Intercept)                                              1.532e+02 112.665
## age_scaled                                               1.554e+02   1.859
## points_scaled                                            5.791e+02   2.057
## abstraction1                                             4.379e+02  23.134
## reward_condition1                                        4.485e+02  -2.400
## age_scaled:points_scaled                                 5.898e+02  -0.070
## age_scaled:abstraction1                                  4.379e+02   0.187
## points_scaled:abstraction1                               4.379e+02   1.715
## age_scaled:reward_condition1                             4.531e+02  -0.953
## points_scaled:reward_condition1                          5.070e+02  -1.065
## abstraction1:reward_condition1                           4.379e+02  -0.041
## age_scaled:points_scaled:abstraction1                    4.379e+02   0.061
## age_scaled:points_scaled:reward_condition1               4.987e+02  -1.443
## age_scaled:abstraction1:reward_condition1                4.379e+02  -0.372
## points_scaled:abstraction1:reward_condition1             4.379e+02   2.366
## age_scaled:points_scaled:abstraction1:reward_condition1  4.379e+02   1.728
##                                                         Pr(>|t|)    
## (Intercept)                                               <2e-16 ***
## age_scaled                                                0.0649 .  
## points_scaled                                             0.0402 *  
## abstraction1                                              <2e-16 ***
## reward_condition1                                         0.0168 *  
## age_scaled:points_scaled                                  0.9442    
## age_scaled:abstraction1                                   0.8515    
## points_scaled:abstraction1                                0.0870 .  
## age_scaled:reward_condition1                              0.3413    
## points_scaled:reward_condition1                           0.2876    
## abstraction1:reward_condition1                            0.9669    
## age_scaled:points_scaled:abstraction1                     0.9515    
## age_scaled:points_scaled:reward_condition1                0.1496    
## age_scaled:abstraction1:reward_condition1                 0.7104    
## points_scaled:abstraction1:reward_condition1              0.0184 *  
## age_scaled:points_scaled:abstraction1:reward_condition1   0.0848 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

	AUC
Predictors	Estimates	SE
age scaled	0.0128	0.0069
points scaled	0.0090	0.0044
abstraction1	0.0603	0.0026
reward condition1	-0.0064	0.0027
age scaled × points scaled	-0.0003	0.0043
age scaled × abstraction1	0.0005	0.0027
points scaled × abstraction1	0.0050	0.0029
age scaled × reward condition1	-0.0026	0.0028
points scaled × reward condition1	-0.0038	0.0036
abstraction1 × reward condition1	-0.0001	0.0026
age scaled × points scaled × abstraction1	0.0002	0.0028
age scaled × points scaled × reward condition1	-0.0048	0.0033
age scaled × abstraction1 × reward condition1	-0.0010	0.0027
points scaled × abstraction1 × reward condition1	0.0070	0.0029
age scaled × points scaled × abstraction1 × reward condition1	0.0048	0.0028
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.62
N _{subject_id}	151
Observations	608
Marginal R² / Conditional R²	0.303 / 0.734

Figure 3B: AUC by performance group and reward condition

Do choice weights relate to memory?

Model: AUC by age, exemplar choice weights, specificity, block condition

	AUC
Predictors	Estimates	SE
age scaled	0.0097	0.0068
beta scaled	0.0180	0.0047
abstraction1	0.0610	0.0025
reward condition1	-0.0052	0.0026
age scaled × beta scaled	0.0159	0.0055
age scaled × abstraction1	0.0038	0.0025
beta scaled × abstraction1	-0.0044	0.0027
age scaled × reward condition1	-0.0003	0.0026
beta scaled × reward condition1	-0.0097	0.0031
abstraction1 × reward condition1	-0.0001	0.0025
age scaled × beta scaled × abstraction1	-0.0041	0.0030
age scaled × beta scaled × reward condition1	-0.0075	0.0033
age scaled × abstraction1 × reward condition1	0.0011	0.0025
beta scaled × abstraction1 × reward condition1	-0.0006	0.0027
age scaled × beta scaled × abstraction1 × reward condition1	0.0044	0.0030
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.62
N _{subject_id}	151
Observations	608
Marginal R² / Conditional R²	0.322 / 0.741

Figure 3C: AUC by exemplar choice weights - model effects

Model: AUC by age, category choice weights, specificity, block condition

	AUC
Predictors	Estimates	SE
age scaled	0.0196	0.0072
beta scaled	0.0000	0.0043
abstraction1	0.0593	0.0029
reward condition1	-0.0076	0.0034
age scaled × beta scaled	-0.0017	0.0042
age scaled × abstraction1	0.0025	0.0029
beta scaled × abstraction1	0.0070	0.0029
age scaled × reward condition1	-0.0029	0.0034
beta scaled × reward condition1	0.0001	0.0036
abstraction1 × reward condition1	-0.0019	0.0029
age scaled × beta scaled × abstraction1	-0.0012	0.0029
age scaled × beta scaled × reward condition1	-0.0070	0.0036
age scaled × abstraction1 × reward condition1	0.0014	0.0029
beta scaled × abstraction1 × reward condition1	0.0022	0.0029
age scaled × beta scaled × abstraction1 × reward condition1	-0.0023	0.0029
Random Effects
σ²	0.00
τ₀₀ _{subject_id}	0.01
ICC	0.64
N _{subject_id}	151
Observations	608
Marginal R² / Conditional R²	0.290 / 0.741

Figure 3C: AUC by category choice weights: model effects

Relations between age and other model parameters

Model: Initial Q values by age

	Estimate	Std. Error	t value	Pr(>\|t\|)
(Intercept)	0.7336	0.09476	7.742	1.38e-12
age_scaled	-0.2009	0.09508	-2.113	0.03625

Fitting linear model: q_init ~ age_scaled
Observations	Residual Std. Error	\(R^2\)	Adjusted \(R^2\)
151	1.164	0.0291	0.02258

Supplementary Figure: Initial Q values by age

Model: Alpha values by age

	Estimate	Std. Error	t value	Pr(>\|t\|)
(Intercept)	-0.6771	0.07922	-8.547	1.39e-14
age_scaled	0.007934	0.07949	0.09982	0.9206

Fitting linear model: alpha ~ age_scaled
Observations	Residual Std. Error	\(R^2\)	Adjusted \(R^2\)
151	0.9735	6.686e-05	-0.006644

Memory Specificity Analyses: Experiment 1

Kate Nussenbaum

11/13/23

Import and process data

Learning

Model: Correct responses by age, trial, block number, and block condition

Figure 2A: Correct response by block condition, stimulus repetition, and age group

Supplementary Figure: Correct response by block condition and block number

Figure 2B: Generalization by block condition, category repetition, age group

Model: Correct response to first appearance of each stimulus

Supplementary Figure: Generalization by trial, reward condition, and block number

Figure 2C: WSLS by age group

Supplementary Figure: WSLS by block number

Model: Category win-stay lose-shift

Stats: WSLS in first 10 trials

Memory

Memory delay stats

Supplementary Figure: AUC values by stimulus repetitions, block condition, age group

Model: AUC values by stimulus repetitions, block condition, age

Figure 3A: AUC values by age group, memory specificity, block condition

Model: AUCs by age, reward condition, memory specificity

Model: AUCs by age, reward condition, foil type, memory delay

Supplementary Figure: AUC values with delay

RL modeling

Choice weights

Figure 2E: Distribution of choice weights

Model: Choice weights by block condition and age

Model: Exemplar choice weights across conditions

Model: Category choice weights across conditions

Relations between choice weights and points earned

Figure 2D: Effect of choice weights on points earned

Relations between learning and memory

Do points earned during learning relate to memory?

Model: AUC by points earned

Figure 3B: AUC by performance group and reward condition

Do choice weights relate to memory?

Model: AUC by age, exemplar choice weights, specificity, block condition

Figure 3C: AUC by exemplar choice weights - model effects

Model: AUC by age, category choice weights, specificity, block condition

Figure 3C: AUC by category choice weights: model effects

Relations between age and other model parameters

Model: Initial Q values by age

Supplementary Figure: Initial Q values by age

Model: Alpha values by age